Research Of Pruning Redundant Rules In Positive And Negative Association Rules

Association rule mining is one of the most important techniques of data mining. But the traditional association rules discovery algorithm produces too many redundant rules,which makes it difficult for users to analyze and makes use of these rules. To facilitate analysis, the number of redundant rules can be reduced significantly by techniques. At present,many researchers have proposed some non-redundant rules algorithm.In this paper,some arisen redundant rules is summarized and discussed .This paper takes corresponding method to prune these redundant rules.(1) An improved non-redundant rules method is presented based on former algorithm.The method shows an important theorem through introducing correlation.If A(?)BC is an effective association rule,then A(?)B(or A(?)C) is an effective association rule iff corrA,B>1(corrA,C >1).This theorm explains that A(?)B(or A(?)C) is redundant for A(?)BC no other than corrA,B>1(corrA,C >1).Otherwise, A(?)B(or A(?)C)is not an effective rule. This paper presents an improved non-redundant rules method based on this theorem. The experiments show that this method can reduce the number of association rules effectively.(2) Negtive association rules mined by PNARC is analyzed in this paper,we found many redundant rules.This paper only discusses three kind of typical negtive association rules,①(?)A(?)(?)B(?)C,②(?)A(?)BC,③(?)A(?)(?)BC. Because these three kind of rules make more redundant rules for them.Therefore,this paper educed three important deduce and proved them. The experiments show that this method can reduce the number of association rules in a certain extent.